The AI arms race
Artificial intelligence (AI) is a technology growing in importance, becoming a battleground between national powers, among whom the U.S. and China.
As a matter of fact, the US government pledged to increased spending on AI next year from 1.2 billion to 1.5 billion. China, on the other, has recently added AI to the list of technologies that they choose to ban from export. Similar to the race to the moon or the nuclear arms race, AI is the technology everyone wants to get their hands on.
After having made the development of AI a national imperative, China’s State Council published a blueprint stating they wanted to create the “world’s primary AI innovation center” by 2030. They aim to turn AI into a $150 billion industry, pushing its application into smart cities and military uses. For this purpose, they planned to build a $2.1 billion AI-focused technology research park. That’s one way to achieve your objectives.
The Quest for Patents
Each empire has its champion. The United States has high hopes in Nvidia, the market leader in graphic processing units (GPU), used for increasing the speed of AI applications. Facing him however, the Chinese contender is called Cambricon Technologies. Although more recent, Cambricon has no less than 4 times as many patents as Nvidia.
Graph #1: Count of AI hardware patents from Cambricon and Nvidia - data provided by Clarivate Analytics PATINFORMATICS, LLC
On a more global scale, China is beating the United States on technology growth: with 58’990 applications filed in 2019 via WIPO’s (World Intellectual Property Organization) Patent Cooperation Treaty System, they finally beat the US which had 57’840 applications in 2019.
Graph #2: AI hardware patenting trends from the US and China - data provided by Clarivate Analytics PATINFORMATICS, LLC
We should note that China generally has and over-estimated number of patents, as they are filing most of them at home, but Cambricon seems to be an exception: even if its Chinese patents were not included, it would still have more than Nvidia as you can see on the graph #1.
Cambricon Technologies has recently raised $368 million in IPO at the Shanghai Stock Exchange STAR market. The company is now valued at over $85 billion.As of the end of 2019, over 75% of the staff at the company was in the R&D department. Thanks to this, they have the largest patent portfolio in the field. Their Cambricon-1M and Cambricon-1H chips are already in more than 100 million devices. Also, their flagship product, the Cambricon-1A is also used as the very first deep learning processor for commercial use.
What are AI chipsets anyways?
Machine learning is the process where a computer receives a large amount of data in order to process it and carry out a specific task such as recognizing an image or understanding speech. For this process, we have neural networks, brain-inspired networks of connected algorithm called neurons which feed data into each other. Passing the data between each other, they will modify the importance of data by giving it weights. At some point, the weight attributed will get very close to the desired value, at which point the system will be deemed to have “learned” how to carry the task it was designed to perform.
Whether you have understood the precedent explanation or not, what you should remember is that this process requires highly performing hardware. The training and implementing of artificial neural networks require vast amounts of data and hardware than needs to function while at the same time consuming as little energy as possible. AI hardware (or chipsets) are designed in a way to accelerate AI software systems and satisfy its specific requirements.
Companies like Cambricon Technologies are not yet profitable, but if they do find a way to turn their many patents into commercial success, things will become interesting. The export ban imposed by the government may slow its expansion, although giving in some field advantage in China.
The war for the AI technology is ongoing. Although the US are making a real effort to keep the pace, China seems to be ahead and will accelerate in the future. For now, Chinese companies are very happy to satisfy the local demand for AI hardware, but at some point, they will have to expand to new markets in order to grow.